Design

Design is the intelligent transformation of indeterminate situations into determinate ones — a process of inquiry that operates not on propositions but on configurations. Where science seeks to describe what is, and engineering seeks to build what works, design seeks to compose what ought to be — not as a moral injunction but as a functional achievement: the creation of forms that resolve problematic situations by reorganizing the relations among their elements.

This definition places design at the intersection of Dewey's pragmatism and systems thinking. Dewey described inquiry as the reconstruction of experience; design is inquiry made material. The designer does not merely think about a problem — they reshape the environment so that the problem dissolves. A well-designed door handle does not remind you to pull; it makes pulling the only natural action. A well-designed institution does not police compliance; it makes compliance the path of least resistance. Design operates on the affordance structure of situations — the possibilities for action that the environment presents — rather than on the intentions of agents within them.

Design as Second-Order Problem Solving

Herbert Simon, in The Sciences of the Artificial (1969), defined design as the core activity of all artificial sciences: every professional who devises courses of action to change existing situations into preferred ones is a designer. This is broader than the visual arts or industrial design. An economist designing a tax policy, a physician designing a treatment protocol, a teacher designing a curriculum — all are engaged in design, because all are transforming indeterminate situations into determinate ones through the reorganization of relations.

The key distinction is that design operates on the problem-situation itself, not merely on the solution. A first-order problem solver selects from available solutions to a well-defined problem. A designer redefines the problem by restructuring the situation so that new solutions become possible. This is why design is often called wicked problem solving: the problem and solution co-evolve, and the act of solving changes the nature of what was being solved.

Design and Feedback Architecture

The design process is a feedback loop structurally identical to the loops in reinforcement learning, control theory, and cybernetics. The designer proposes a configuration (action), the environment responds (users behave in unanticipated ways, materials fail in unexpected modes, contexts shift), and the designer revises (iteration). The difference is that design feedback is slower, messier, and more interpretively laden than algorithmic feedback. A gradient descent step is a mathematical operation; a design iteration is a hermeneutic one — the designer must interpret what the environment's response means, not merely measure it.

This interpretive dimension connects design to hermeneutics and phenomenology. The designer is not a detached optimizer but an embodied interpreter situated within the problem-space. Haraway's situated knowledges apply here: design knowledge is perspectival, partial, and accountable. A design that works for one community may fail for another not because the engineering is wrong but because the affordance structure assumes a body, a culture, a history that not all users share.

Design Failures as Model Failures

Design failures are rarely engineering failures. The bridge that collapses is an engineering failure. The app that no one uses is a design failure — the affordance structure was misaligned with the users' actual practices. The Therac-25 radiation accidents were engineering failures (software bugs). The Facebook News Feed algorithm's amplification of outrage is a design failure: the reward function (engagement) was misspecified relative to the social good, and the design optimized the metric rather than the value.

These failures are instances of Goodhart's law at the level of environmental structure. When a design metric becomes a target — clicks, engagement, efficiency, throughput — the design ceases to serve its original purpose. The difference from pure optimization is that design failures reshape the environment. A misoptimized stock portfolio loses money. A misdesigned social media platform reshapes political discourse. The feedback loop of design is not merely consequential. It is world-making: bad design does not just fail; it creates new problems.

Design, Institutions, and Collective Behavior

Institutional design is the application of design thinking to social systems. Constitutions, market regulations, organizational structures, and algorithmic governance systems are all designs — configurations that constrain and enable collective behavior. The mechanism design tradition in economics treats institutions as algorithms and asks how to design rules that induce desired equilibria. The constitutional political economy tradition asks how to design rules for making rules, recognizing that institutional designers are themselves situated within institutions.

The connection to collective behavior is direct. A well-designed institution does not merely aggregate individual preferences. It shapes the space of possible preferences and actions. A voting system does not just count votes; it determines whether strategic voting is rational, whether minority voices are heard, whether compromise is incentivized. The design of the institution is a phase transition in the collective dynamics: small changes in rules can produce large changes in emergent social patterns. This is why institutional design is not merely applied game theory. It is applied complex systems theory: the designer is not optimizing a function but tuning a dynamical system toward a desired basin of attraction.

The Ethics of World-Making

Design is inherently normative. Every design decision encodes a judgment about what users should do, what they should value, and what they should become. A well-designed sidewalk encourages walking. A well-designed highway encourages driving. A well-designed default option in a pension program encourages saving. These are not neutral interventions. They are nudges — subtle restructurings of the choice architecture that make certain behaviors more likely without mandating them.

The ethical dimension is not an add-on. It is built into the definition. To design is to decide what the world ought to be like, and to materialize that decision in configurations that constrain the possible. The designer's responsibility is therefore not merely to make things that work but to make things that work well — where 'well' is not a technical standard but an ethical one. This connects design to virtue ethics (what kind of person does this design cultivate?), care ethics (whose needs does this design serve and whose does it ignore?), and justice (how does this design distribute affordances across populations?).

The most consequential design question of the present moment is not 'how do we make AI that works?' but 'how do we design the human-AI relationship so that human flourishing is enhanced rather than displaced?' This is not a technical question. It is a design question, and its answer requires not better algorithms but better configurations: institutional, educational, social, and material configurations that situate AI within human practices rather than replacing those practices.